During evolution, there has been an expansion in the diversity of histone acetyltransferase Gcn5-containing complexes. All Gcn5 complexes share Sgf29 and Ada3 subunits, which together with their respective Ada2 homolog, enable nucleosomal HAT activity. In metazoans, including Drosophila, there are two homologs of the Gcn5-binding protein Ada2, Ada2a and Ada2b, which nucleate formation of the Ada2a-containing (ATAC) or SAGA transcription coactivator complexes, respectively. In Drosophila, Ada2b has two splice isoforms that differ in their C-terminal regions but share a common N-terminal region containing the zinc finger-like ZZ, SANT, and two of the three ADA box domains. Thus, we sought to determine whether these Ada2b isoforms were both required in SAGA, or alternatively, if like Ada2a, each Ada2b isoform nucleates formation of a distinct Gcn5-containing complex. To identify proteins that interacted specifically with each isoform, we epitope-tagged the short Ada2b-PA and long Ada2b-PB isoforms at either their N- or C-terminal end, performed tandem affinity purification, and analyzed the co-purifying proteins by MudPIT. Affinity purification followed by mass spectrometry was also performed for other Gcn5 subunits (Spt3, Spt20 and Sgf29) as well as two proteins specifically identified in the short Ada2b-PA purifications, Chiffon and Cdc7. Control purifications from cells expressing a non-specific tagged bait protein, or from cells lacking tagged protein were also analyzed by MudPIT. In brief, TCA-precipitated protein eluates were urea-denatured, reduced, alkylated, and digested with endoproteinase LysC followed by trypsin. The peptide mixtures were loaded onto microcapillary fused silica columns (100um i.d.), packed with C18 reverse phase (Aqua; Phenomenx), SCX (Luna; Phenomenex) and C18-RP, placed in-line with an Agilent 11000 quaternary pump, and analyzed by a 10-step MudPIT on linear ion traps. MS/MS datasets were searched using SEQUEST (v. 27.9) against a non-redudant drosophila protein database (NCBI, 2011-04-11) containing 18,425 non-redundant fly proteins and 177 usual contaminants. To estimate false discover rates (FDRs), the amino acid sequence of each non-redundant protein was randomized (18602 shuffled proteins). Peptide/spectrum matches were sorted and selected using DTASelect with the following criteria set: spectra/peptide matches were retained only if they had a DeltCn of at least 0.8, and minimum XCorr of 1.8 for singly, 2.0 for doubly, and 3.0 for triply charged spectra. Additionally, the peptides had to be minimum 7 amino acids in length and fully tryptic.